Steel Distributor Turns Emailed Orders into ERP Entries
Quick Summary
Oiki, a family-run stainless steel producer and distributor in Parma, Italy, uses an AI agent to read customer order emails, pull out product codes and quantities, and push validated orders toward its ERP. Order processing dropped from 30 minutes to 12.
The Use Case
Most of Oiki's orders arrive as free-form emails. Sales reps had to read each one, extract the product codes and quantities by hand, look them up in internal systems and key the order in. That slowed responses and capped how much volume the team could handle. Working with IBM, Oiki deployed a language model to read the emails and an agent built on watsonx Orchestrate to interpret intent, check the extracted data against business rules and trigger the next step. Reps work with the agent inside Microsoft Teams, so nothing about their day-to-day tools changed.
Real Example
- A customer emails an order listing several alloys, sizes and quantities
- The rep drops the email into the agent's interface in Teams
- The model identifies the intent (new order versus a quote request) and extracts every product code and quantity
- The agent validates the data against Oiki's business rules and flags anything that does not match the catalog
- A clean order triggers the downstream steps toward the ERP; an exception goes back to the rep with the problem highlighted
- The rep confirms and moves to the next email
Why It Matters
- Order processing time fell 60%, from 30 minutes to 12 per order
- Bid management time fell from 300 minutes a day to 120, freeing 3 hours daily
- Oiki expects another 45% gain once the ERP integration is complete, without adding headcount